论文
22AI 评分(两次独立评分的平均)

Artificial Analysis 对比 Parallel Search 与 OpenAI 搜索 token 消耗

原标题:Total Tokens per Task counts the input, reasoning, and answer tokens the model uses per task. Parallel Search (advanced, objective) uses ...

Artificial Analysis@ArtificialAnlys发布时间 (北京时间)

Artificial Analysis 统计了各搜索工具每任务消耗的总 token(含输入、推理与回答),Parallel Search(advanced, objective)约 82k,在 26 个工具中第二低,仅次于 OpenAI Web Search(medium)的 41k。

中文译文尚未完成,先为你展示原文。

Total Tokens per Task counts the input, reasoning, and answer tokens the model uses per task. Parallel Search (advanced, objective) uses about 82k tokens per task, the 2nd lowest of 26, behind OpenAI Web Search (medium) at 41k, which is an integrated first-party search tool. Parallel Search (fast, objective) uses about 116k, comparable to Keenable Search (pro) at 118k and Keenable Search (realtime) at 122k.

In our August runs, advanced used about 168k and fast about 292k.

Nearly all of these are input tokens, which include the search results returned to the model.

数据与训练评测基准#评测/基准#搜索#数据/训练

同一件事的其他报道3 篇